Upload
kyle-botabara
View
186
Download
2
Tags:
Embed Size (px)
DESCRIPTION
Rubiks Cube Group Theory
Citation preview
Group Theory and the Rubik’s Cube
Janet Chen
A Note to the Reader
These notes are based on a 2-week course that I taught for high school students at the Texas State HonorsSummer Math Camp. All of the students in my class had taken elementary number theory at the camp, soI have assumed in these notes that readers are familiar with the integers mod n as well as the units mod n.
Because one goal of this class was a complete understanding of the Rubik’s cube, I have tried to use notationthat makes discussing the Rubik’s cube as easy as possible. For example, I have chosen to use right groupactions rather than left group actions.
2
Introduction
Here is some notation that will be used throughout.
Z the set of integers . . . ,−3,−2,−1, 0, 1, 2, 3, . . .N the set of positive integers 1, 2, 3, . . .Q the set of rational numbers (fractions)R the set of real numbers
Z/nZ the set of integers mod n(Z/nZ)× the set of units mod n
The goal of these notes is to give an introduction to the subject of group theory, which is a branch of themathematical area called algebra (or sometimes abstract algebra). You probably think of algebra as addition,multiplication, solving quadratic equations, and so on. Abstract algebra deals with all of this but, as thename suggests, in a much more abstract way! Rather than looking at a specific operation (like addition) ona specific set (like the set of real numbers, or the set of integers), abstract algebra is algebra done withoutreally specifying what the operation or set is. This may be the first math you’ve encountered in whichobjects other than numbers are really studied!
A secondary goal of this class is to solve the Rubik’s cube. We will both develop methods for solving theRubik’s cube and prove (using group theory!) that our methods always enable us to solve the cube.
References
Douglas Hofstadter wrote an excellent introduction to the Rubik’s cube in the March 1981 issue of ScientificAmerican. There are several books about the Rubik’s cube; my favorite is Inside Rubik’s Cube and Beyondby Christoph Bandelow. David Singmaster, who developed much of the usual notation for the Rubik’s cube,also has a book called Notes on Rubik’s ’Magic Cube,’ which I have not seen.
For an introduction to group theory, I recommend Abstract Algebra by I. N. Herstein. This is a wonderfulbook with wonderful exercises (and if you are new to group theory, you should do lots of the exercises). Ifyou have some familiarity with group theory and want a good reference book, I recommend Abstract Algebraby David S. Dummit and Richard M. Foote.
3
1. Functions
To understand the Rubik’s cube properly, we first need to talk about some different properties of functions.
Definition 1.1. A function or map f from a domain D to a range R (we write f : D → R) is a rule whichassigns to each element x ∈ D a unique element y ∈ R. We write f(x) = y. We say that y is the image of xand that x is a preimage of y. Note that an element in D has exactly one image, but an element of R mayhave 0, 1, or more than 1 preimage.
Example 1.2. We can define a function f : R→ R by f(x) = x2. If x is any real number, its image is thereal number x2. On the other hand, if y is a positive real number, it has two preimages,
√y and −√y. The
real number 0 has a single preimage, 0; negative numbers have no preimages. ❖
Functions will provide important examples of groups later on; we will also use functions to “translate”information from one group to another.
Definition 1.3. A function f : D → R is called one-to-one if x1 6= x2 implies f(x1) 6= f(x2) for x1, x2 ∈ D.That is, each element of R has at most one preimage.
Example 1.4. Consider the function f : Z→ R defined by f(x) = x+ 1. This function is one-to-one since,if x1 6= x2, then x1 + 1 6= x2 + 1. If x ∈ R is an integer, then it has a single preimage (namely, x − 1). Ifx ∈ R is not an integer, then it has no preimage.
The function f : Z→ Z defined by f(x) = x2 is not one-to-one, since f(1) = f(−1) but 1 6= −1. Here, 1 hastwo preimages, 1 and −1. ❖
Definition 1.5. A function f : D → R is called onto if, for every y ∈ R, there exists x ∈ D such thatf(x) = y. Equivalently, every element of R has at least one preimage.
Example 1.6. The function f : Z→ R defined by f(x) = x+ 1 is not onto since non-integers do not havepreimages. However, the function f : Z→ Z defined by f(x) = x+ 1 is onto.
The function f : Z→ Z defined by f(x) = x2 is not onto because there is no x ∈ Z such that f(x) = 2. ❖
Exercise 1.7. Can you find a . . .
1. . . . function which is neither one-to-one nor onto?2. . . . function which is one-to-one but not onto?3. . . . function which is onto but not one-to-one?4. . . . function which is both one-to-one and onto?
Definition 1.8. A function f : D → R is called a bijection if it is both one-to-one and onto. Equivalently,every element of R has exactly one preimage.
Example 1.9. The function f : Z→ Z defined by f(x) = x+ 1 is a bijection. ❖
Example 1.10. If S is any set, then we can define a map f : S → S by f(x) = x for all x ∈ S. This mapis called the identity map, and it is a bijection. ❖
Definition 1.11. If f : S1 → S2 and g : S2 → S3, then we can define a new function f ◦ g : S1 → S3 by(f ◦ g)(x) = g(f(x)). The operation ◦ is called composition.
Remark 1.12. One usually writes (g ◦ f)(x) = g(f(x)) rather than (f ◦ g)(x) = f(g(x)). However, as longas we are consistent, the choice does not make a big difference. We are using this convention because itmatches the convention usually used for the Rubik’s cube.
4
Exercises
1. Which of the following functions are one-to-one? Which are onto?
(a) f : Z→ Z defined by f(x) = x2 + 1.(b) f : N→ N defined by f(x) = x2 + 1.(c) f : Z→ Z defined by f(x) = 3x+ 1.(d) f : R→ R defined by f(x) = 3x+ 1.
2. Suppose f1 : S1 → S2 and f2 : S2 → S3 are one-to-one. Prove that f1 ◦ f2 is one-to-one.
3. Suppose f1 : S1 → S2 and f2 : S2 → S3 are onto. Prove that f1 ◦ f2 is onto.
4. Let f1 : S1 → S2, f2 : S2 → S3, and f3 : S3 → S4. Prove that f1 ◦ (f2 ◦ f3) = (f1 ◦ f2) ◦ f3.
5. Let S be a set.
(a) Prove that there exists a function e : S → S such that e ◦ f = f and f ◦ e = f for all bijectionsf : S → S. Prove that e is a bijection. S.
(b) Prove that, for every bijection f : S → S, there exists a bijection g : S → S such that f ◦ g = eand g ◦ f = e.
6. If f : D → R is a bijection and D is a finite set with n elements, prove that R is also a finite set withn elements.
5
2. Groups
Example 2.1. To get an idea of what groups are all about, let’s start by looking at two familiar sets.
First, consider the integers mod 4. Remember that Z/4Z is a set with 4 elements: 0, 1, 2, and 3. One ofthe first things you learned in modular arithmetic was how to add numbers mod n. Let’s write an additiontable for Z/4Z.
+ 0 1 2 30 0 1 2 31 1 2 3 02 2 3 0 13 3 0 1 2
Now, we’re going to rewrite the addition table in a way that might seem pretty pointless; we’re just goingto use the symbol ∗ instead of + for addition, and we’ll write e = 0, a = 1, b = 2, and c = 3. Then, ouraddition table looks like
∗ e a b ce e a b ca a b c eb b c e ac c e a b
Let’s do the same thing for (Z/5Z)×, the set of units mod 5. The units mod 5 are 1, 2, 3, and 4. If you addtwo units, you don’t necessarily get another unit; for example, 1 + 4 = 0, and 0 is not a unit. However, ifyou multiply two units, you always get a unit. So, we can write down a multiplication table for (Z/5Z)×.Here it is:
· 1 2 4 31 1 2 4 32 2 4 3 14 4 3 1 23 3 1 2 4
Again, we’re going to rewrite this using new symbols. Let ∗ mean multiplication, and let e = 1, a = 2, b = 4,and c = 3. Then, the multiplication table for (Z/5Z)× looks like
∗ e a b ce e a b ca a b c eb b c e ac c e a b
Notice that this is exactly the same as the table for addition on Z/4Z!
Why is it interesting that we get the same tables in these two different situations? Well, this enables us totranslate algebraic statements about addition of elements of Z/4Z into statements about multiplication ofelements of (Z/5Z)×. For example, the equation x + x = 0 in Z/4Z has two solutions, x = 0 and x = 2.With our alternate set of symbols, this is the same as saying that the equation x ∗ x = e has solutions x = eand x = b. If we translate this to (Z/5Z)×, this says that the solutions of x · x = 1 in (Z/5Z)× are x = 1and x = 4. That is, 1 and 4 are the square roots of 1 in (Z/5Z)×, which is exactly right!
In mathematical language, we say that Z/4Z with addition and (Z/5Z)× with multiplication are “isomorphicgroups.” The word “isomorphic” means roughly that they have the same algebraic structure; we’ll get intothis later. For now, let’s just see what a “group” is. ❖
6
Definition 2.2. A group (G, ∗) consists of a set G and an operation ∗ such that:
1. G is closed under ∗. That is, if a, b ∈ G, then a ∗ b ∈ G.
Examples:
• Z/4Z is closed under +; after all, we wrote down the addition table, which tells us how to addany two elements of Z/4Z and get another element of Z/4Z. Similarly, (Z/5Z)× is closed undermultiplication.
• Z is closed under +: if a, b ∈ Z, then a+ b ∈ Z. Similarly, Z is also closed under −.• R is closed under multiplication: if we multiply two real numbers, we get a real number.• The set of negative numbers is not closed under multiplication: if we multiply two negative num-
bers, we get a positive number.
2. ∗ is associative. That is, for any a, b, c ∈ G, a ∗ (b ∗ c) = (a ∗ b) ∗ c.
Examples:
• Addition and multiplication are associative.• Subtraction is not associative because a− (b− c) 6= (a− b)− c.
3. There is an “identity element” e ∈ G which satisfies g = e ∗ g = g ∗ e for all g ∈ G.
Examples:
• For (Z/4Z,+), 0 is an identity element because g = 0 + g = g + 0 for any g ∈ Z/4Z. For((Z/5Z)×, ·), 1 is an identity element because g = 1 · g = g · 1 for any g ∈ (Z/5Z)×.
• For (Z,+), 0 is an identity element because g = 0 + g = g + 0 for any g ∈ Z.• For (R, ·), 1 is an identity element because g = 1 · g = g · 1 for any g ∈ R.
4. Inverses exist; that is, for any g ∈ G, there exists an element h ∈ G such that g ∗ h = h ∗ g = e. (h iscalled an inverse of g.)
Examples:
• Using the addition table for Z/4Z, we can find inverses of all the elements of Z/4Z. For instance,we can see from the table that 1 + 3 = 3 + 1 = 0, so 3 is the inverse of 1. Similarly, since thetable for (Z/5Z)× is identical, all elements of (Z/5Z)× have inverses.
• For (Z,+), the inverse of n ∈ Z is −n because n+ (−n) = (−n) + n = 0.• For (R, ·), not every element has an inverse — namely, 0 does not have an inverse. However, ifx 6= 0, then 1
x is an inverse of x because x · 1x = 1
x · x = 1.
Example 2.3.
1. (Z/4Z,+) and ((Z/5Z)×, ·) are groups. In fact, as we said earlier, these should be thought of as the“same” group, but we won’t go into this until later.
2. (Z,+) is a group. However, (Z,−) is not a group because subtraction is not associative.3. (R, ·) is not a group since 0 does not have an inverse under multiplication. However, (R− {0}, ·) is a
group.4. The set of negative numbers is not closed under multiplication, so the set of negative numbers with
multiplication is not a group.5. We can construct a group (G, ∗) where G is a set with just one element. Since G must have an identity
element, we will call this single element e. To define the group operation ∗, we just need to say whate ∗ e is. There is only one choice since G has only one element: e ∗ e must be e. This defines a groupwhich is called the trivial group. As you might guess, the trivial group isn’t very interesting.
6. Soon, we will see how to make the moves of a Rubik’s cube into a group!
7
❖
The examples of groups we have seen so far all have another special property: for every g, h ∈ G, g∗h = h∗g;that is, the ∗ operation is commutative. This is not true of all groups. If it is true of (G, ∗), we say (G, ∗) isabelian. We will soon see examples of nonabelian groups.
Now, we will prove two important properties of groups.
Lemma 2.4. A group has exactly one identity element.
Proof. Let (G, ∗) be a group, and suppose e and e′ are identity elements of G (we know that G has at leastone identity element by the definition of a group). Then, e ∗ e′ = e since e′ is an identity element. On theother hand, e ∗ e′ = e′ since e is an identity element. Therefore, e = e′ because both are equal to e ∗ e′.
Lemma 2.5. If (G, ∗) is a group, then each g ∈ G has exactly one inverse.
Proof. Let g ∈ G, and suppose g1, g2 are inverses of G (we know there is at least one by the definition of agroup); that is, g ∗ g1 = g1 ∗ g = e and g ∗ g2 = g2 ∗ g = e. By associativity, (g1 ∗ g) ∗ g2 = g1 ∗ (g ∗ g2).Since g1 is an inverse of g, (g1 ∗ g) ∗ g2 = e ∗ g2 = g2. Since g2 is an inverse of g, g1 ∗ (g ∗ g2) = g1 ∗ e = g1.Therefore, g2 = g1.
In general, we write the unique inverse of g as g−1. However, if we know that the group operation is addition,then we write the inverse of g as −g.
Exercises
1. Which of the following are groups? Prove your answer.
(a) ({±1}, ·)(b) (S,+), where S is the set of non-negative integers {0, 1, 2, 3, . . .}(c) (2Z,+), where 2Z is the set of even integers(d) (Z, ·)(e) (Z, ?) where a?b is defined to be a+ b− 1.(f) (Q− {0}, ·)(g) (R,★) where a★ b is defined to be (a− 1)(b− 1).(h) (G, ◦) where G is the set of bijections from some set S to itself and ◦ denotes composition of
functions.
2. Let (G, ∗) be a group and a, b, c ∈ G. Prove:
(a) If a ∗ b = a ∗ c, then b = c.(b) If b ∗ a = c ∗ a, then b = c.
That is, we can cancel in groups.
3. If (G, ∗) is a group and g ∈ G, prove that (g−1)−1 = g.
4. If (G, ∗) is a group and g, h ∈ G such that g ∗ h = e, prove that h ∗ g = e. (That is, if g ∗ h = e, then his the inverse of g and g is the inverse of h.)
5. If (G, ∗) is a group and g, h ∈ G such that g ∗ h = h, prove that g is the identity element of G.
6. Let (G, ∗) be a finite group; that is, (G, ∗) is a group and G has finitely many elements. Let g ∈ G.Prove that there exists a positive integer n such that gn = e (here, gn means g ∗g ∗ · · · ∗g with n copiesof g). The smallest such integer n is called the order of g.
7. Find the order of 5 in (Z/25Z,+). Find the order of 2 in ((Z/17Z)×, ·).
8
8. If (G, ∗) is a group in which g ∗ g = e for all g ∈ G, show that (G, ∗) is abelian.
9. If (G, ∗) is a group where G has 4 elements, show that (G, ∗) is abelian.
10. Let (G, ∗) be a finite group. Prove that there is a positive integer n such that gn = e for all g ∈ G.(This is different from Problem 6 in that you need to show that the same n works for all g ∈ G.)
11. Let G be a set and ∗ be an operation on G such that the following four properties are satisfied:
(a) G is closed under ∗.(b) ∗ is associative.(c) There exists e ∈ G such that g ∗ e = g for all g ∈ G. (We call e a “right identity.”)(d) For each g ∈ G, there exists h ∈ G such that g ∗ h = e. (We call h a “right inverse” of g.)
Prove that (G, ∗) is a group.
9
3. The Rubik’s Cube and Subgroups
3.1. Cube notation
The Rubik’s cube is composed of 27 small cubes, which are typically called “cubies.” 26 of these cubies arevisible (if you take your cube apart, you’ll find that the 27th cubie doesn’t actually exist). When workingwith the Rubik’s cube, it’s helpful to have a systematic way of referring to the individual cubies. Althoughit seems natural to use the colors of a cubie, it is actually more useful to have names which describe thelocations of the cubies. The cubies in the corners are called, appropriately enough, “corner cubies.” Eachcorner cubie has 3 visible faces, and there are 8 corner cubies. The cubies with two visible faces are called“edge” cubies; there are 12 edge cubies. Finally, the cubies with a single visible face are called “centercubies,” and there are 6 center cubies.
Now, let’s name the 6 faces of the Rubik’s cube. Following the notation developed by David Singmaster,we will call them right (r), left (l), up (u), down (d), front (f), and back (b). The advantage of this namingscheme is that each face can be referred to by a single letter.
To name a corner cubie, we simply list its visible faces in clockwise order. For instance, the cubie in theupper, right, front corner is written urf. Of course, we could also call this cubie rfu or fur. Sometimes, wewill care which face is listed first; in these times, we will talk about “oriented cubies.” That is, the orientedcubies urf, rfu, and fur are different. In other situations, we won’t care which face is listed first; in thesecases, we will talk about “unoriented cubies.” That is, the unoriented cubies urf, rfu, and fur are the same.
Similarly, to name edge and center cubies, we will just list the visible faces of the cubies. For instance, thecubie in the center of the front face is just called f, because its only visible face lies on the front of the cube.
We will also frequently talk about “cubicles.” These are labeled the same way as cubies, but they describethe space in which the cubie lives. Thus, if the Rubik’s cube is in the start configuration (that is, the Rubik’scube is solved), then each cubie lives in the cubicle of the same name (the urf cubie lives in the urf cubicle,the f cubie lives in the f cubicle, and so on). If you rotate a face of the Rubik’s cube, the cubicles don’tmove, but the cubies do. Notice, however, that when you rotate a face of the Rubik’s cube, all center cubiesstay in their cubicles.
Finally, we want to give names to some moves of the Rubik’s cube. The most basic move one can do is torotate a single face. We will let R denote a clockwise rotation of the right face (looking at the right face,turn it 90◦ clockwise). Similarly, we will use the capital letters L, U, D, F, and B to denote clockwise twistsof the corresponding faces. More generally, we will call any sequence of these 6 face twists a “move” of theRubik’s cube. For instance, rotating the right face counterclockwise is a move which is the same as doing Rthree times. Later in this lecture, we will describe a notation for these more complicated moves.
A couple of things are immediately clear. First, we already observed that the 6 basic moves keep the centercubies in their cubicles. Since any move is a sequence of these 6 basic moves, that means that every moveof the Rubik’s cube keeps the center cubies in their cubicles (for a formal proof, see the example afterProposition 4.9). Also, any move of the Rubik’s cube puts corner cubies in corner cubicles and edge cubiesin edge cubicles; it is impossible for a corner cubie to ever live in an edge cubicle or for an edge cubie to livein a corner cubicle. Using these two facts, we can start to figure out how many possible configurations theRubik’s cube has. Let’s look, for instance, at the urf cubicle. Theoretically, any of the 8 corner cubies couldreside in this cubicle. That leaves 7 corner cubies that could reside in the urb cubicle, 6 for the next cornercubicle, and so on. Therefore, there are 8 · 7 · 6 · 5 · 4 · 3 · 2 · 1 = 8! possible positionings of the corner cubies.Notice that a corner cubie can fit into its cubicle in 3 different ways. For instance, if a red, white, and bluecubie lies in the urf cubicle, either the red, white, or blue face could lie in the u face of the cubicle (and thisdetermines where the other 2 faces lie). Since there are 8 corner cubies and each can lie in its cubicle in 3different ways, there are 38 different ways the corner cubies could be oriented. Therefore, there are 38 · 8!possible configurations of the corner cubies. Similarly, since there are 12 edge cubies, there are 12! positionsof the edge cubies; each edge cubie has 2 possible orientations, giving 212 possible orientations. So, there are
10
212 · 12! possible configurations of the edge cubies, giving a total of 212388!12! possible configurations of theRubik’s cube. (This number is about 5.19× 1020, or 519 quintillion!)
Although these configurations are theoretically possible, that doesn’t mean that these configurations couldreally occur. We will say that a configuration of the Rubik’s cube is valid if it can be achieved by a seriesof moves from the starting configuration. It turns out that some of the theoretically possible configurationswe have counted are actually not valid. Therefore, we have two goals:
1. Demonstrate that some configurations are not valid.2. Find a set of moves that can take us from any valid configuration back to the start configuration.
3.2. Making the Rubik’s Cube into a Group
We can make the set of moves of the Rubik’s cube into a group, which we will denote (G, ∗). The elementsof G will be all possible moves of the Rubik’s cube (for example, one possible move is a clockwise turn ofthe top face followed by a counterclockwise turn of the right face). Two moves will be considered the sameif they result in the same configuration of the cube (for instance, twisting a face clockwise by 180◦ is thesame as twisting the same face counterclockwise by 180◦). The group operation will be defined like this: ifM1 and M2 are two moves, then M1 ∗M2 is the move where you first do M1 and then do M2.
Why is this a group? We just need to show the 4 properties in [PS 2, #11].
• G is certainly closed under ∗ since, if M1 and M2 are moves, M1 ∗M2 is a move as well.
• If we let e be the “empty” move (that is, a move which does not change the configuration of the Rubik’scube at all), then M ∗ e means “first do M , then do nothing.” This is certainly the same as just doingM , so M ∗ e = M . So, (G, ∗) has a right identity.
• If M is a move, we can reverse the steps of the move to get a move M ′. Then, the move M ∗M ′ means“first do M , then reverse all the steps of M .” This is the same as doing nothing, so M ∗M ′ == e, soM ′ is the inverse of M . Therefore, every element of G has a right inverse.
• Finally, we must show that ∗ is associative. Remember that a move can be defined by the change inconfiguration it causes. In particular, a move is determined by the position and orientation it putseach cubie in.
If C is an oriented cubie, we will write M(C) for the oriented cubicle that C ends up in after we applythe move M , with the faces of M(C) written in the same order as the faces of C. That is, the firstface of C should end up in the first face of M(C), and so on. For example, the move D puts the urcubie in the br cubicle, with the u face of the cubie lying in the b face of the cubicle and the r face ofthe cubie lying in the r face of the cubicle. Thus, we write D(ur) = br.
First, let’s investigate what a sequence of two moves does to the cubie. If M1 and M2 are two moves,then M1 ∗M2 is the move where we first do M1 and then do M2. The move M1 moves C to the cubicleM1(C); the move M2 then moves it to M2(M1(C)). Therefore, (M1 ∗M2)(C) = M2(M1(C)).
To show that ∗ is associative, we need to show that (M1 ∗M2) ∗M3 = M1 ∗ (M2 ∗M3) for any movesM1, M2, and M3. This is the same as showing that (M1 ∗M2) ∗M3 and M1 ∗ (M2 ∗M3) do the samething to every cubie. That is, we want to show that [(M1 ∗M2) ∗M3](C) = [M1 ∗ (M2 ∗M3)](C) forany cubie C. We know from our above calculation that [(M1 ∗M2) ∗M3](C) = M3([M1 ∗M2](C)) =M3(M2(M1(C))). On the other hand, [M1 ∗ (M2 ∗M3)](C) = (M2 ∗M3)(M1(C)) = M3(M2(M1(C))).So, (M1 ∗M2) ∗M3 = M1 ∗ (M2 ∗M3). Thus, ∗ is associative.
Therefore, (G, ∗) is indeed a group.
11
3.3. Subgroups
We calculated that there are around 519 quintillion possible configurations of the Rubik’s cube (althoughthese are not all valid). Trying to understand such a large number of configurations is no easy task! It ishelpful to restrict the problem; for instance, instead of looking at all possible moves of the Rubik’s cube, wemight start out by looking at the moves which only involve twists of the down and right faces.
This is a general philosophy in group theory: to understand a group G, we should try to understand smallpieces of it.
Definition 3.1. A nonempty subset H of a group (G, ∗) is called a subgroup of G if (H, ∗) is a group.
The advantage of studying subgroups is that they may be much smaller and, hence, simpler; however, theystill have algebraic structure.
Example 3.2. A group is always a subgroup of itself. Also, the trivial group is a subgroup of any group.However, these subgroups aren’t too interesting! ❖
Example 3.3. The set of even integers is a subgroup of (Z,+): after all, the even integers are certainly asubset of Z, and we know from [PS 2, #1c] that (2Z,+) is a group. ❖
The following lemma often makes it easier to check if a subset is actually a subgroup.
Lemma 3.4. Let (G, ∗) be a group. A nonempty subset H of G is a subgroup of (G, ∗) iff, for every a, b ∈ H,a ∗ b−1 ∈ H.
Proof. First, suppose H is a subgroup. If b ∈ H, then b−1 ∈ H since (H, ∗) is a group. So, if a ∈ H as well,then a ∗ b−1 ∈ H.
Conversely, suppose that, for every a, b ∈ H, a ∗ b−1 ∈ H.
• First, notice that ∗ is associative since (G, ∗) is a group.• Let a ∈ H. Then, e = a ∗ a−1, so e ∈ H.• Let b ∈ H, Then b−1 = e ∗ b−1 ∈ H, so inverses exist in H.• Let a, b ∈ H. By the previous step, b−1 ∈ H, so a ∗ (b−1)−1 = a ∗ b ∈ H. Thus, H is closed under ∗.
Therefore, (H, ∗) is a group, which means that H is a subgroup of G.
3.4. Simplifying Group Notation
It is common practice to write group operations as multiplication; that is, we write gh rather than g ∗h, andwe call this the “product” of g and h. The statement “let G be a group” really means that G is a groupunder some operation which will be written as multiplication. We will also often write the identity elementof G as 1 rather than e. Finally, we will use standard exponential notation, so g2 means gg, g3 means ggg,and so on.
In particular, we will do this with (G, ∗). That is, from now on, we will just call this group G, and we willwrite the operation as multiplication. For instance, DR means the move D followed by the move R. The movewhich twists the right face counterclockwise by 90◦ is the same as a move twisting the right face clockwisethree times, so we can write this move as R3.
Exercises
1. If G is a group and g, h ∈ G, write (gh)−1 in terms of g−1 and h−1.
12
2. If A,B are subgroups of a group G, prove that A ∩B is a subgroup of G.
3. Let G be a group and Z(G) = {z ∈ G : zx = xz for all x ∈ G}. (This notation means that Z(G) is theset of z ∈ G such that zx = xz for all x ∈ G.) Prove that Z(G) is a subgroup of G. Z(G) is called thecenter of G. If G is abelian, what is Z(G)?
4. Remember that G is the group of moves of the Rubik’s cube. Prove that this group is not abelian.(Hint: pick two moves M1 and M2 and look at their commutator [M1,M2], which is defined to beM1M2M
−11 M−1
2 .)
5. Let C1 and C2 be two different unoriented corner cubies, and let C ′1 and C ′2 be two different unorientedcorner cubicles. Prove that there is a move of the Rubik’s cube which sends C1 to C ′1 and C2 to C ′2.Since we are talking about unoriented cubies and cubicles, we only care about the positions of thecubies, not their orientations. (For example, if C1 = dbr, C2 = urf, C ′1 = dlb, and C ′2 = urf, then themove D sends C1 to C ′1 and C2 to C ′2.)
6. Let G be a group and S be a subset of G. Let H be the set of all elements of G which can be writtenas a finite product of elements of S and their inverses; that is, H consists of all elements of the forms1 · · · sn where each si is either an element of S or the inverse of an element of S. Prove that H is asubgroup of G. We call H the subgroup of G generated by S and write H = 〈S〉. It is the smallestsubgroup of G containing S (do you see why?).
7. If G is an abelian group, show that {a2 : a ∈ G} is a subgroup of G. Which part of your proof failswhen G is not abelian?
8. Find all subgroups of (Z,+).
9. Let (G, ∗) be a group and H be a nonempty finite subset of G closed under ∗. Prove that H is asubgroup of G. Is the statement still true if H is not required to be finite?
10. Try scrambling your Rubik’s cube. How many cubies can you put in the right position? The rightorientation?
11. Brainstorm a bit about what kind of strategy you should use to solve the Rubik’s cube. Which cubiesshould you try to solve first? What kind of moves should you look for?
13
4. Generators
Let G be a group and S be a subset of G. In [PS 3, #6], we defined the subgroup 〈S〉. Now, we will givesome properties of 〈S〉. First, let’s look at the special case when 〈S〉 is just G.
Definition 4.1. Let G be a group and S be a subset of G. We say that S generates G or that S is a set ofgenerators of G if G = 〈S〉; that is, every element of G can be written as a finite product (under the groupoperation) of elements of S and their inverses.
Example 4.2. Every element of Z can be written as a finite sum of 1’s or −1’s, so Z is generated by {1}.That is, Z = 〈1〉. For the same reason, Z = 〈−1〉. Of course, it’s also true that Z = 〈1, 2〉. In general, thereare many possible sets of generators of a group. ❖
Example 4.3. Every element of Z/4Z can be written as a finite sum of 1’s, so Z/4Z = 〈1〉.
Even though Z = 〈1〉 and Z/4Z = 〈1〉, Z and Z/4Z are not equal! 〈S〉 only makes sense in the context of agiven group. ❖
Example 4.4. Every element of G can be written as a finite sequence of turns of the Rubik’s cube, soG = 〈D,U, L,R,F,B〉. ❖
You might think of generators as being the “core” of the group; since every element of the group can bewritten in terms of the generators, knowledge about the generators can often be translated into knowledgeabout the whole group. We will make this more precise soon.
Definition 4.5. A group G is cyclic if there exists g ∈ G such that G = 〈g〉.
Example 4.6. Z and Z/4Z are cyclic. ❖
For finite groups, we can even relax the definition of generators slightly by leaving out the inverses of S. Toprove this, we need a lemma.
Lemma 4.7. Let G be a finite group and g ∈ G. Then, g−1 = gn for some n ∈ N.
Proof. If g = e, then there is nothing to show. So, suppose g 6= e. By [PS 2, #6], there exists a positiveinteger m such that gm = e. Since g 6= e, m 6= 1, so m > 1. Let n = m− 1 ∈ N. Then, ggn = gm = e, so gn
is the inverse of g by [PS 2, #4].
Lemma 4.8. Let G be a finite group and S be a subset of G. Then, G = 〈S〉 iff every element of G can bewritten as a finite product of elements of S. (That is, the inverses of S are not necessary.)
Proof. If every element of G can be written as a finite product of elements of S, then it is clear that G = 〈S〉.
Conversely, suppose G = 〈S〉. This means that every element of G can be written as a finite product s1 · · · snwhere each si is either in S or the inverse of an element of S. The basic point of the proof is that the inverseof an element of S can also be written as a product of elements of S by the previous lemma. To make thiscompletely rigorous, we will use induction on n.
Suppose n = 1. Either s1 ∈ S or s−11 ∈ S. If s1 ∈ S, then s1 is written as a product of a single element of
S. If s−11 ∈ S, then s1 can be written as a finite product of elements of S by Lemma 4.7. So, the base case
is true.
Now, suppose the statement is true for all natural numbers smaller than n; we want to show that s1 · · · sncan be written as a finite product of elements of S. By the induction hypothesis, s1 . . . , sn−1 and sn can bothbe written as finite products of elements of S. Therefore, their product s1 · · · sn certainly can as well.
Now, we will see how to translate properties of generators to the whole group.
14
Proposition 4.9. Let G be a finite group and S be a subset of G. Suppose the following two conditions aresatisfied.
1. Every element of S satisfies some property P .2. If g ∈ G and h ∈ G both satisfy the property P , then gh satisfies the property P as well.
Then, every element of 〈S〉 satisfies P .
Before we prove this, let’s see how it might be used.
Example 4.10. Let S = {D,U, L,R,F,B} ⊂ G. Then, every M ∈ S satisfies the property “M keeps allcenter cubies in their cubicles.” If M1,M2 ∈ G are such that they keep all center cubies in their cubicles,then M1M2 certainly keeps all center cubies in their cubicles. Since G = 〈S〉, the proposition says that everyelement of G keeps the center cubies in their cubicles.
This proposition is extremely useful for the Rubik’s cube because it means we frequently only need tounderstand properties of the 6 basic moves rather than all 5× 1020 possible moves. ❖
Now, let’s prove Proposition 4.9.
Proof. By Lemma 4.8, any element of 〈S〉 can be written as s1 · · · sn where n ∈ N and each si is an elementof S. We will prove the proposition by induction on n.
If n = 1, then s1 ∈ S satisfies property P by hypothesis.
Suppose inductively that s1 · · · sn−1 satisfies property P . Then, the product (s1 · · · sn−1)sn is a product oftwo elements satisfying property P , so it satisfies property P as well.
Exercises
1. If H is a subgroup of a group G, prove that 〈H〉 = H.
2. If A is a subset of B, prove that 〈A〉 is a subgroup of 〈B〉. Is the converse true?
3. Let G be a nontrivial group (remember that the trivial group is the group with only one element, so anontrivial group is a group with at least 2 elements). Suppose that the only subgroups of G are G and{1}. Prove that G is cyclic and finite, and prove that the number of elements in G is a prime number.
4. Prove that any subgroup of a cyclic group is cyclic.
5. Remember that G is the group of moves of the Rubik’s cube, D ∈ G is a clockwise twist of the downface, and R ∈ G is a clockwise twist of the right face. Find the order of D, R, and DR (remember, DRis the move where you first do D, then R; for the definition of order, see [PS 2, #6]).
6. A group G is finitely generated if there exists a finite subset S of G such that G = 〈S〉.
(a) Prove that every finitely generated subgroup of (Q,+) is cyclic.(b) Prove that (Q,+) is not finitely generated.
15
5. The Symmetric Group
When we found the number of possible configurations of the Rubik’s cube, we used the fact that any movesends corner cubies to corner cubicles to deduce that there are 8! possible ways to position the corner cubies.In this section, we will lay down the mathematical foundation needed to understand these possibilities.
Rather than just looking at configurations of 8 cubies, we’ll look at configurations of any n objects. We’llcall these objects 1, 2, . . . , n, although these names are arbitrary. We can think of arranging these objects asputting them into n slots. If we number the slots 1, 2, . . . , n, then we can define a function σ : {1, 2, . . . , n} →{1, 2, . . . , n} by letting σ(i) be the number put into slot i.
Example 5.1. We can put the objects 1, 2, 3 in the order 3 1 2. Here, 3 is in the first slot, 1 is in thesecond slot, and 2 is in the third slot. So, this ordering corresponds to the function σ : {1, 2, 3} → {1, 2, 3}defined by σ(1) = 3, σ(2) = 1, and σ(3) = 2. ❖
What can we say about σ?
Lemma 5.2. σ is a bijection.
Proof. Suppose x 6= y. Since a number cannot be in more than one slot, if x 6= y, slots x and y must containdifferent numbers. That is, σ(x) 6= σ(y). Therefore, σ is one-to-one.
Any number y ∈ {1, 2, . . . , n} must lie in some slot, say slot x. Then, σ(x) = y. So, σ is onto.
On the other hand, if σ : {1, . . . , n} → {1, . . . , n} is a bijection, then σ defines an arrangement of the nobjects: just put object σ(i) in slot i. So, the set of possible arrangements is really the same as the set ofbijections {1, . . . , n} → {1, . . . , n}. Therefore, instead of studying possible arrangements, we can study thesebijections.
Definition 5.3. The symmetric group on n letters is the set of bijections from {1, 2, . . . , n} to {1, 2, . . . , n},with the operation of composition. We write this group as Sn.
Note that Sn is a group by [PS 2, #1h].
Let’s do an example to make sure that the group operation is clear.
Example 5.4. Let σ, τ ∈ S3 be defined by σ(1) = 3, σ(2) = 1, σ(3) = 2, τ(1) = 1, τ(2) = 3, and τ(3) = 2.Then, (στ)(1) = τ(3) = 2, (στ)(2) = τ(1) = 1, and (στ)(3) = τ(2) = 3. ❖
5.1. Disjoint Cycle Decomposition
There is a more compact way of writing elements of the symmetric group; this is best explained by anexample.
Example 5.5. Consider σ ∈ S12 defined by
σ(1) = 12 σ(2) = 4 σ(3) = 5 σ(4) = 2 σ(5) = 6 σ(6) = 9σ(7) = 7 σ(8) = 3 σ(9) = 10 σ(10) = 1 σ(11) = 11 σ(12) = 8
We will write ”i 7→ j” (“i maps to j”) to mean σ(i) = j. Then,
1 7→ 12, 12 7→ 8, 8 7→ 3, 3 7→ 5, 5 7→ 6, 6 7→ 9, 9 7→ 10, 10 7→ 12 7→ 4, 4 7→ 27 7→ 711 7→ 11
16
This data tells us what σ does to each number, so it defines σ. As shorthand, we write
σ = (1 12 8 3 5 6 9 10) (2 4) (7) (11).
Here, (1 12 8 3 5 6 9 10), (2 4), (7), and (10) are called cycles. When writing the disjoint cycle de-composition, we leave out the cycles with just one number, so the disjoint cycle decomposition of σ isσ = (1 12 8 3 5 6 9 10) (2 4). ❖
Now, let’s actually define what a cycle is.
Definition 5.6. The cycle (i1 i2 · · · ik) is the element τ ∈ Sn defined by τ(i1) = i2, τ(i2) = i3, . . . , τ(ik−1) =ik, τ(ik) = i1 and τ(j) = j if j 6= ir for any r. The length of this cycle is k, and the support of the cycleis the set {i1, . . . , ik} of numbers which appear in the cycle. The support is denoted by supp τ . A cycle oflength k is also called a k-cycle.
Definition 5.7. Two cycles σ and τ are disjoint if they have no numbers in common; that is, suppσ ∩supp τ = ∅.
Lemma 5.8. Let σ, τ ∈ Sn be cycles. If σ and τ are disjoint, then στ = τσ.
Proof. Let i ∈ {1, . . . , n}. Since suppσ ∩ supp τ = ∅, there are only two possibilities:
• i 6∈ suppσ and i 6∈ supp τ . In this case, σ(i) = i and τ(i) = i, so (σ ◦ τ)(i) = τ(i) = i and(τ ◦ σ)(i) = σ(i) = i.
• Otherwise, i is in the support of exactly one of σ and τ . We may suppose without loss of generality thati 6∈ suppσ and i ∈ supp τ . Then, σ(i) = i, so (σ ◦ τ)(i) = τ(i). On the other hand, (τ ◦σ)(i) = σ(τ(i)).Now, since τ(i) ∈ supp τ and supp τ ∩ suppσ = ∅, τ(i) 6∈ suppσ. Therefore, σ(τ(i)) = τ(i). So, weagain have (στ)(i) = (τσ)(i).
Therefore, (στ)(i) = (τσ)(i) = i for all i, which shows that στ = τσ.
Any σ ∈ Sn can be written as a product (under the group operation, which is composition) of disjoint cycles.This product is called the disjoint cycle decomposition of σ. In our example, we gave a method for findingthe disjoint cycle decomposition of a permutation.
We write the identity permutation as 1.
Example 5.9. S2 consists of two permutations, 1 and (1 2). ❖
Example 5.10. Let σ, τ ∈ S6 be defined by
σ(1) = 3 σ(2) = 5 σ(3) = 4 σ(4) = 1 σ(5) = 2 σ(6) = 6τ(1) = 5 τ(2) = 4 τ(3) = 3 τ(4) = 2 τ(5) = 1 τ(6) = 6
In cycle notation, σ = (1 3 4)(2 5) and τ = (1 5)(2 4). Then, στ = (1 3 2) (4 5) and τσ = (1 2)(3 4 5). Wecan also easily compute σ2 = (1 4 3) and τ2 = 1. ❖
Definition 5.11. If σ ∈ Sn is the product of disjoint cycles of lengths n1, . . . , nr (including its 1-cycles),then the integers n1, . . . , nr are called the cycle type of σ.
5.2. Rubik’s Cube
We can write each move of the Rubik’s cube using a slightly modified cycle notation. We want to describewhat happens to each oriented cubie; that is, we want to describe where each cubie moves and where eachface of the cubie moves. For example, if we unfold the cube and draw the down face, it looks like
17
f f fl d d d rl d d d rl d d d r
b b b
If we rotate this face clockwise by 90◦ (that is, we apply the move D), then the down face looks like
l l lb d d d fb d d d fb d d d f
r r r
So, D(dlf) = dfr because the dlf cubie now lives in the dfr cubicle (with the d face of the cubie lying in the dface of the cubicle, the l face of the cubie lying in the f face of the cubicle, and the f face of the cubie lyingin the r face of the cubicle). Similarly, D(dfr) = drb, D(drb) = dbl, and D(dbl) = dlf. If we do the same thingfor the edge cubies, we find D = (dlf dfr drb dbl)(df dr db dl).
Example 5.12. Check that the disjoint cycle decomposition of R is (rfu rub rbd rdf)(ru rb rd rf). ❖
Exercises
1. Let σ, τ ∈ S5 be defined as follows.
σ(1) = 3 σ(2) = 4 σ(3) = 5 σ(4) = 2 σ(5) = 1τ(1) = 5 τ(2) = 3 τ(3) = 2 τ(4) = 4 τ(5) = 1
Find the cycle decompositions of each of the following permutations: σ, τ , σ2, and τ2σ.
2. Let σ = (1 12 8 10 4)(2 13)(5 11 7)(6 9). Find σ2 and σ3. What is the order of σ?
3. Suppose σ is a permutation with cycle type n1, . . . , nr. What is the order of σ?
4. Let σ = (1 2) and τ = (2 3). Find στ and τσ.
(a) What is the order of (1 2)(2 3)? Does this agree with your answer to [PS 5, #3]?(b) For what n is Sn abelian?
5. Write all the elements of S3, the symmetric group on 3 elements, using disjoint cycle notation.
6. Find all subgroups of S3.
7. Remember that D, U, L, R, F, and B are defined to be clockwise twists of the down, up, left, right,front, and back faces, respectively. We showed in class that D has disjoint cycle decomposition(dlf dfr drb dbl)(df dr db dl) and that R has disjoint cycle decomposition (rfu rub rbd rdf)(ru rb rd rf).Write U, L, F, and B as products of disjoint cycles.
8. Let a1, . . . , am be distinct elements of {1, . . . , n}, and let σ = (a1 a2)(a1 a3)(a1 a4) · · · (a1 am). Findthe disjoint cycle decomposition of σ.
9. Show that Sn is generated by the set of 2-cycles in Sn.
10. Let τ ∈ Sn, and let a1, . . . , ak) be distinct elements of {1, . . . , n}. Prove that τ−1(a1 a2 · · · ak)τ =(τ(a1) τ(a2) · · · τ(ak)).
18
11. Let σ, σ′ ∈ S6 be defined by
σ = (1 3 2)(4 6)σ′ = (2 5 4)(1 6)
Find τ ∈ S6 such that σ′ = τ−1στ .
12. Let σ, σ′ ∈ Sn. Prove that σ and σ′ have the same cycle type iff σ′ = τ−1στ for some τ ∈ Sn.
Note: Any element of the form τ−1στ is called a conjugate of σ, so this says that the conjugates of σare exactly the elements of Sn with the same cycle type as σ.
13. Let H be the subgroup of G generated by D2 and R2; that is, H = 〈D2,R2〉. How many elements doesH have? Which of these elements might be helpful for solving the Rubik’s cube?
14. Let σ, τ ∈ Sn. Suppose suppσ∩supp τ = {x}; that is, x is the only number in {1, . . . , n} which appearsin the cycle decomposition of both σ and τ . We define the commutator of σ and τ to be στσ−1τ−1,and we write this as [σ, τ ]. Show that supp[σ, τ ] has at most 3 elements.
19
6. Configurations of the Rubik’s Cube
As we said already, a configuration of the Rubik’s cube is determined by four pieces of data:
• the positions of the corner cubies• the positions of the edge cubies• the orientations of the corner cubies• the orientations of the edge cubies
The first can be described by an element σ of S8 (i.e., the element of S8 which moves the corner cubiesfrom their start positions to the new positions). The second can be described by an element τ of S12. Now,we will see how to understand the third and fourth. The basic idea is to fix a “starting orientation” and asystematic way of writing down how a given orientation differs from this starting orientation. Ths is mostlyjust a matter of notation.
We’ll start with the corner cubies. Each corner cubie has 3 possible orientations, and we will number theseorientations 0, 1, and 2. Let’s explain what these numbers mean. Imagine that your Rubik’s cube is in thestart configuration. We are going to write a number on one face of each corner cubicle, as follows. Write:
1 on the u face of the ufl cubicle2 on the u face of the urf cubicle3 on the u face of the ubr cubicle4 on the u face of the ulb cubicle5 on the d face of the dbl cubicle6 on the d face of the dlf cubicle7 on the d face of the dfr cubicle8 on the d face of the drb cubicle
So, each corner cubicle now has exactly one numbered face. Each corner cubie thus has one face lying in anumbered cubicle face. Label this cubie face 0. Going around the cubie clockwise, label the next face 1, andthen label the final face 2.
Example 6.1. If we look straight at the down face and unfold the cube, the cubie faces look like this.
f f fl d d d rl d d d rl d d d r
b b b
So, the cubicle numberings that we can see look like this:
6 7
5 8
Therefore, the cubie labels look like
2 11 0 0 2
2 0 0 11 2
20
❖
Now, each face of each corner cubie has a number on it. If the Rubik’s cube is in any configuration, wewill describe the orientations of the corner cubies like this: for any i between 1 and 8, find the cubicle facelabeled i; let xi be the number of the cubie face living in this cubicle face. We write x for the ordered 8-tuple(x1, . . . , x8). Notice that we can think of each xi as counting the number of clockwise twists the cubie i isaway from having its 0 face in the numbered face of the cubicle. But a cubie that is 3 twists away is orientedthe same way as a cubie that is 0 twists away. Thus, we should think of the xi as being elements of Z/3Z.So, x is an 8-tuple of elements of Z/3Z; we write x ∈ (Z/3Z)8.
Example 6.2. If the Rubik’s cube is in the start configuration, each xi is 0. We also write x = 0 to meanthat each xi is 0. ❖
Example 6.3. Let’s see what the xi are after we apply the move R to a cube in the start configuration. Inthe start configuration, the right hand face looks like this:
u u uf r r r bf r r r bf r r r b
d d d
The cubicle numbers on this face are
2 3
7 8
Therefore, the labeling of the corner cubies looks like this:
0 02 1 2 1
1 2 1 20 0
If we rotate the right face of the cube by 90◦, then the cubie faces look like
1 20 2 1 0
0 1 2 02 1
The cubies on the left face are unaffected by R, so x1 = 0, x4 = 0, x5 = 0, and x6 = 0. Now, we can seefrom our diagrams that x2 = 1, x3 = 2, x7 = 2, and x8 = 1. So, x = (0, 1, 2, 0, 0, 0, 2, 1). ❖
We can do the same thing for the edge cubies. First, we label the edge cubicles as follows. Write:
21
1 on the u face of the ub cubicle2 on the u face of the ur cubicle3 on the u face of the uf cubicle4 on the u face of the ul cubicle5 on the b face of the lb cubicle6 on the b face of the rb cubicle7 on the f face of the rf cubicle8 on the f face of the lf cubicle9 on the d face of the db cubicle10 on the d face of the dr cubicle11 on the d face of the df cubicle12 on the d face of the dl cubicle
Each edge cubie now has a face lying in a numbered cubicle face; label this cubie face 0, and label the otherface of the cubie 1. Then, let yi be the number of the cubie face in the cubicle face numbered i. This definesy ∈ (Z/2Z)12.
Thus, any configuration of the Rubik’s cube can be described by σ ∈ S8, τ ∈ S12, x ∈ (Z/3Z)8, andy ∈ (Z/2Z)12. So, we will write configurations of the Rubik’s cube as ordered 4-tuples (σ, τ, x, y).
Example 6.4. The start configuration is (1, 1, 0, 0). ❖
Example 6.5. Pretend that your cube is in the start configuration. Let (σ, τ, x, y) be the configuration ofthe cube after we do the move [D,R], which is defined to be DRD−1R−1. We will write down σ, τ , x, and y.
We showed in class that D = (dlf dfr drb dbl)(df dr db dl) and R = (rfu rub rbd rdf)(ru rb rd rf). Therefore,D−1 = (dbl drb dfr dlf)(dl db dr df) and R−1 = (rdf rbd rub rfu)(rf rd rb ru). So,
[D,R] = (dlf dfr lfd frd fdl rdf)(drb bru bdr ubr rbd rub)(df dr br) (6.1)
(When writing this down, be very careful with the orientations.)
Remember that τ is an element of S12; we think of it as a bijection from the set of 12 unoriented edgecubies to the set of 12 edge cubicles. It is defined like this: if C is an unoriented edge cubie in the startconfiguration, then τ(C) is the unoriented edge cubicle where C is living in the current configuration. Likeany element of S12, τ can be written in disjoint cycle notation.
In this particular example, [D,R] moves cubie df to cubicle dr, cubie dr to cubicle br, and cubie br to cubicledf. Therefore, τ = (df dr br).
Similarly, we think of σ as a bijection from the set of 8 unoriented corner cubies to the set of 8 unorientedcorner cubicles. To find σ, we must figure out what [D,R] does to the positions of the corner cubies. Observethat [D,R] switches the positions of the cubies dfl and dfr, and it also switches the positions of drb and bru.Therefore, σ = (drb bru)(dfl dfr).
Recall that we defined x as follows. When the cube was in the start configuration, we numbered 8 cubiclefaces like this:
1 on the u face of the ufl cubicle2 on the u face of the urf cubicle3 on the u face of the ubr cubicle4 on the u face of the ulb cubicle5 on the d face of the dbl cubicle6 on the d face of the dlf cubicle7 on the d face of the dfr cubicle8 on the d face of the drb cubicle
22
We numbered each of the corresponding cubie faces 0. Starting from 0 on a corner cubie, we went clockwiseand labeled the other two faces 1 and 2. (For example, the u face of the ufl cubie is labeled 0, so the f faceis 1 and the l face is 2.) Now that the cube is no longer in the start configuration, we define xi to be thecubie face number in cubicle face i.
In the start position, all of the numbered cubicle faces have cubie faces numbered 0. Since the move [D,R]does not affect cubies ufl, urf, ulb, or dbl, x1, x2, x4, and x5 must be 0. To find x3, we want to see whichcubie face is in the u face of the ubr cubicle. We can see from (6.1) that [D,R] puts the b face of the drb cubiethere; by our numbering scheme, the b face of the drb cubie is numbered 2; therefore, x3 = 2. Similarly,x6 = 2, x7 = 0, and x8 = 2. Therefore, the ordered 8-tuple x is (0, 0, 2, 0, 0, 2, 0, 2).
Similarly, to define y, we first numbered 12 edge cubicle faces (when the cube was in the start configuration):
1 on the u face of the ub cubicle2 on the u face of the ur cubicle3 on the u face of the uf cubicle4 on the u face of the ul cubicle5 on the b face of the lb cubicle6 on the b face of the rb cubicle7 on the f face of the rf cubicle8 on the f face of the lf cubicle9 on the d face of the db cubicle10 on the d face of the dr cubicle11 on the d face of the df cubicle12 on the d face of the dl cubicle
Then, we labeled the corresponding cubie faces 0 and the other edge cubie faces 1. Finally, we defined y tobe the 12-tuple (y1, . . . , y12) where yi is the number in edge cubicle face i.
Since [D,R] only affects the edge cubies df, dr, and br, we know right away that y11, y10, and y6 are the onlyyi that may be nonzero. Since [D,R] puts the b face of the br cubicle in the d face of the df cubicle, y11 = 0.Similarly, y10 and y6 are both 0. So, y = (0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0).
You might wonder why we bother to define configurations of the Rubik’s cube this way, since it seems mucheasier to get the information directly from (6.1). The main reason is that writing σ, τ , x, and y separatelyallows us to recognize patterns more easily (and prove them!). ❖
Exercises
1. Suppose your Rubik’s cube is in the configuration (σ, τ, x, y). If you apply the move D to the cube, itends up in a new configuration (σ′, τ ′, x′, y′). Write x′ and y′ in terms of x and y. Do the same for themoves U, L, R, F, and B. Do you notice any patterns?
2. Write the commutator [D,R] in disjoint cycle notation (be careful to keep track of the orientations ofthe cubies). What is the order of [D,R]? Can you find . . .
(a) . . . a move which fixes the positions (but not necessarily orientations) of the back corner cubiesand one front corner cubie?
(b) . . . a move which leaves 6 corner cubies fixed and switches the other 2?
Try using these moves to put all the corner cubies in the right positions (but not necessarily orienta-tions). What other useful moves can you get from [D,R]?
3. (a) In the subgroup 〈D,R〉 of G, find a move which changes the orientations (but not positions) oftwo corner cubies without affecting any other corner cubies. Your move may do anything to edge
23
cubies. (Hint: compute a few elements of 〈D,R〉 and think about which powers of these elementsare simplest. Then try to put some of these simple things together.)
(b) Find a move which changes the orientations of the cubies dbr and ufl but does not affect any othercorner cubies.
24
7. Group Homomorphisms
In Example 2.1, we said that Z/4Z and (Z/5Z)× should be thought of as the same group. After all, theyboth have 4 elements, and addition in Z/4Z behaves exactly the same way as multiplication in (Z/5Z)×. Inorder to understand this, we really need a way to translate behavior from one group to another.
Let’s take another look at Example 2.1. We observed in that example that the addition table for Z/4Z andthe multiplication table for (Z/5Z)× were really the same. Essentially, if we replaced 0, 1, 2, and 3 in theaddition table for Z/4Z by 1, 2, 4, 3 (respectively), then we got the the multiplication table for (Z/5Z)×.Another way of thinking about this is that we really defined a function f : Z/4Z → (Z/5Z)× by f(0) = 1,f(1) = 2, f(2) = 4, and f(3) = 3. This function had a special property, though, which was that it translatedthe addition table for Z/4Z to the multiplication table for (Z/5Z)×.
How can we express this? Well, the idea is that, for every a, b ∈ Z/4Z, the term a + b in the additiontable for Z/4Z should correspond to the term f(a)f(b) in the multiplication table for (Z/5Z)×. That is,f(a + b) = f(a)f(b) for all a, b ∈ Z/4Z. This condition expresses the fact that f “translates” the additiontable for Z/4Z to the multiplication table for (Z/5Z)×.
We can generalize this condition to any pair of groups.
Definition 7.1. Let (G, ∗) and (H,★) be two groups. A homomorphism from G to H is a map φ : G→ Hsuch that φ(a ∗ b) = φ(a) ★φ(b) for all a, b ∈ G.
Example 7.2. We can define a map φ : Z/4Z → (Z/5Z)× by taking φ(0) = 1, φ(1) = 2, φ(2) = 4, andφ(3) = 3. Using the addition table for Z/4Z and the multiplication table for (Z/5Z)×, we can check thatφ(a+ b) = φ(a)φ(b) for all a, b ∈ Z/4Z. Alternatively, observe that φ(x) = 2x for all x, so φ(a+ b) = 2a+b =2a2b = φ(a)φ(b). ❖
Example 7.3. We can define a map φcorner : G → S8 as follows. Any move in G certainly rearrangesthe corner cubies somehow; thus, it defines a permutation of the 8 unoriented corner cubies. That is, anyM ∈ G defines some permtuation σ ∈ S8. Let φcorner(M) = σ. That is, φcorner(M) is the element ofS8 which describes what M does to the unoriented corner cubies. For example, we know that [D,R] hasdisjoint cycle decomposition (dlf dfr lfd frd fdl rdf)(drb bru bdr ubr rbd rub)(df dr br). Therefore, φcorner(M) =(dlf dfr)(drb bru).
Similarly, we can define a homomorphism φedge : G→ S12 by letting φedge(M) be the element of S12 whichdescribes what M does to the 12 unoriented edge cubies. For example, φedge([D,R]) = (df dr br).
Finally, we can define a “cube” homomorphism φcube : G → S20, which describes permutations of the 20unoriented edge and corner cubies. For example, φcorner([D,R]) = (dlf dfr)(drb bru)(df dr br). ❖
Exercises
1. Let σ ∈ S5 be defined by σ(1) = 2, σ(2) = 4, σ(3) = 1, σ(4) = 5, and σ(5) = 3. Write σ as a productof 2-cycles in at least 3 different ways (this is possible by [PS 5, #9]). Do you notice any patterns inthe ways you have written σ? How many 2-cycles did you use?
2. Let φ : (Z,+) → (R− {0}, ·) be defined by φ(x) = 2x. Prove that φ is a homomorphism.
3. Find all homomorphisms φ : (Z/4Z,+) → ((Z/5Z)×, ·).
4. (a) Find a move M ∈ G which switches the positions of the urf and bdl cubies without affecting anyother corner cubies. What is φcorner(M)?
Hint: Start with the move you found in [PS 6, #2b].
25
(b) Let C1 and C2 be any two distinct unoriented corner cubies. Prove that there is some moveM ∈ G which switches the positions of C1 and C2 without affecting any other corner cubies.(Even if you didn’t find the move in [PS 7, #4a], assume that such a move exists, and you shouldbe able to do this proof!) What is φcorner(M)?
5. Let φ : (G, ∗) → (H,★) be a homomorphism.
(a) Let 1G be the identity element of G and 1H be the identity element of H. Prove that φ(1G) = 1H .
(b) Prove that φ(g−1) = φ(g)−1 for all g ∈ G. (First make sure you understand exactly what thismeans!)
6. Let φ : (G, ∗) → (H,★) be a homomorphism. The image of φ is defined to be the set im φ = {φ(g) :g ∈ G}. Prove that im φ is a subgroup of H.
26
8. The Sign Homomorphism
By [PS 5, #9], Sn is generated by the 2-cycles in Sn. That is, any permutation in Sn can be written asa finite product of 2-cycles. However, any given permutation of Sn can be written as a finite product of2-cycles in infinitely many ways, so it seems like there is not much we can say about this product.
Some permutations in Sn can be written as a product of an even number of 2-cycles; we call these evenpermutations. Other permutations in Sn can be written as a product of an odd number of 2-cycles; we callthese odd permutations. So far, there seems to be no reason that a permutation could not be both even andodd. However, it is in fact true that a permutation is either even or odd, but not both.
Unfortunately, a direct proof of this fact is rather messy; instead, we will give a proof which is relativelysimple but uses an indirect trick.
Fix n, and let p(x1, . . . , xn) be a polynomial in the n variables x1, . . . , xn.
Example 8.1. If n = 1, p(x1) is a polynomial in the variable x1; that is, p(x1) looks like amxm1 +am−1xm−11 +
· · ·+ a0. So, p(x1) is a sum of terms that look like axi1.
If n = 2, then p(x1, x2) is a sum of terms that look like axi1xj2.
In general, p(x1, . . . , xn) is a sum of terms that look like axi11 xi22 · · ·xinn . ❖
If σ ∈ Sn, let pσ be the polynomial defined by (pσ)(x1, . . . , xn) = p(xσ(1), . . . , xσ(n)). That is, we simplyreplace xi by xσ(i).
Example 8.2. Suppose n = 4, p(x1, x2, x3, x4) = x31 + x2x3 + x1x4, and σ ∈ S4 has cycle decomposition
σ = (1 2 3). Then, (pσ)(x1, x2, x3, x4) = x3σ(1) + xσ(2)xσ(3) + xσ(1)xσ(4) = x3
2 + x3x1 + x2x4. ❖
Lemma 8.3. For any σ, τ ∈ Sn, (pσ)τ = pστ .
This statement is easy to misinterpret, so before we give a proof, let’s do an example.
Example 8.4. As in the previous example, let n = 4, p(x1, x2, x3, x4) = x31 + x2x3 + x1x4, and σ = (1 2 3).
Let τ = (1 3) (2 4). We know that pσ = x32 + x3x1 + x2x4, so (pσ)τ = x3
τ(2) + xτ(3)xτ(1) + xτ(2)xτ(4) =x3
4 + x1x3 + x4x2. On the other hand, στ = (1 4 2), so pστ = x3(στ)(1) + x(στ)(2)x(στ)(3) + x(στ)(1)x(στ)(4) =
x34 + x1x3 + x4x2. ❖
Now, we will prove Lemma 8.3.
Proof. By definition, (pσ)(x1, . . . , xn) = p(xσ(1), . . . , xσ(n)), so [(pσ)τ ](x1, . . . , xn) = p(xτ(σ(1)), . . . , xτ(σ(n))).Now, τ(σ(i)) = (στ)(i), so [(pσ)τ ](x1, . . . , xn) = p(x(στ)(1), . . . , x(στ)(n)) = (pστ )(x1, . . . , xn).
To prove our assertion about even and odd permutations, we will apply Lemma 8.3 to a specific polynomial,namely
∆ =∏
1≤i<j≤n
(xi − xj).
Example 8.5. If n = 3, ∆ = (x1 − x2)(x1 − x3)(x2 − x3). ❖
Lemma 8.6. For any σ ∈ Sn, ∆σ = ±∆.
Example 8.7. If σ = (1 3 2), then ∆σ = (x3 − x1)(x3 − x2)(x1 − x2) = (x1 − x2)(x1 − x3)(x2 − x3) = ∆.
On the other hand, if σ = (1 2), then ∆σ = (x2 − x1)(x2 − x3)(x1 − x3) = −∆. ❖
Now, we will prove Lemma 8.6. As you might guess from the examples, the idea is to match terms of ∆with terms of ∆σ. That is, for each term xi− xj of the product for ∆, either xi− xj or its negative appears
27
in the product for ∆σ.
Proof. By definition,∆ =
∏1≤i<j≤n
(xi − xj),
so∆σ =
∏1≤i<j≤n
(xσ(i) − xσ(j)).
In order to show ∆σ = ±∆, we must show two things. First, for each i and j with 1 ≤ i < j ≤ n, we mustshow that either xσ(i) − xσ(j) or its negative appears in ∆; that is, either xσ(i) − xσ(j) or its negative hasthe form xk − x` with 1 ≤ k < l ≤ n. Secondly, we must show that, for each i and j with 1 ≤ i < j ≤ n,either xi − xj or its negative appears in ∆σ. Since ∆ and ∆σ have the same number of terms, these twostatements together prove that the terms of ∆ and ∆σ match up.
To prove the first statement, all we need to show is that either σ(i) < σ(j) or σ(j) < σ(i); equivalently, weneed to show that σ(i) 6= σ(j) if 1 ≤ i < j ≤ n. This is true because σ is one-to-one and i 6= j.
To prove the second statement, we need to show that either xi − xj or its negative can be written asxσ(k) − xσ(`) with 1 ≤ k < ` ≤ n. Since σ ∈ Sn, σ−1 ∈ Sn; in particular, σ−1 is also a bijection. Since i 6= j,σ−1(i) 6= σ−1(j). Let k be the smaller of σ−1(i) and σ−1(j), and let ` be the larger. Then, 1 ≤ k < ` ≤ n,and xi − xj is either xσ(k) − xσ(`) or its negative.
By Lemma 8.6, we can define a map ε : Sn → {±1} by σ∆ = ε(σ)∆. By Lemma 8.3, ∆στ = (∆σ)τ =[ε(σ)∆]τ = ε(σ)∆τ = ε(σ)ε(τ)∆. Therefore, ε(στ) = ε(σ)ε(τ). So, ε is a homomorphism. We call it the signhomomorphism.
We claimed at the beginning that ε(σ) had something to do with the number of 2-cycles in a productdecomposition of σ. Now, we’ll prove this.
Theorem 8.8. If σ is a 2-cycle, then ε(σ) = −1.
Proof. First, let σ = (1 2). We can write
∆ =∏
1≤i<j≤n
(xi − xj).
Now, let’s write the terms where i = 1 or i = 2 separately. Then,
∆ =∏
1<j≤n
(x1 − xj)∏
2<j≤n
(x2 − xj)∏
3≤i<j≤n
(xi − xj)
= (x1 − x2)∏
2<j≤n
(x1 − xj)∏
2<j≤n
(x2 − xj)∏
3≤i<j≤n
(xi − xj)
Therefore,
∆σ = (xσ(1) − xσ(2))∏
2<j≤n
(xσ(1) − xσ(j))∏
2<j≤n
(xσ(2) − xσ(j))∏
3≤i<j≤n
(xσ(i)−xσ(j))
= (x2 − x1)∏
2<j≤n
(x2 − xj)∏
2<j≤n
(x1 − xj)∏
3≤i<j≤n
(xi − xj)
= −∆
Thus, we have proved the statement for σ = (1 2).
We could generalize the above argument to any 2-cycle, but there is an easier way! Let σ be any 2-cycle. By[PS 5, #12], σ is conjugate to (1 2). That is, σ = τ(1 2)τ−1 for some τ ∈ Sn. Since ε is a homomorphism,ε(σ) = ε(τ)ε(1 2)ε(τ)−1 = ε(1 2) = −1.
28
Since ε is multiplicative, if ε(σ) = 1, then σ must be a product of an even number of 2-cycles. Similarly, ifε(σ) = −1, then σ must be a product of an odd number of 2-cycles. So, σ is even iff ε(σ) = 1, and σ is oddiff ε(σ) = −1.
Exercises
1. Let φ : (G, ∗) → (H,★) be a homomorphism. If S is a subset of im φ, prove that 〈S〉 is a subgroup ofim φ.
2. Prove that the homomorphism φcorner : G → S8 is onto (equivalently, im φcorner is S8). What doesthis tell you about the possible positions of the corner cubies?
3. Suppose your Rubik’s cube is in the configuration (σ, τ, x, y). If you apply the move M ∈ G to theRubik’s cube, it ends up in a new configuration (σ′, τ ′, x′, y′). Prove that σ′ = σφcorner(M) andτ ′ = τφedge(M).
4. Let (σ, τ, x, y) be a configuration of the Rubik’s cube. Prove that there is a move M ∈ G which putsall of the corner cubies in the correct positions.
5. Let φ : G → H be a homomorphism and G′ be a subgroup of G. Define a map φ′ : G′ → H byφ′(g) = φ(g). Prove that φ′ is a homomorphism. We call this homomorphism the restriction of φ toG′ and write φ′ = φ|G′ .
6. Find all homomorphisms φ : Z→ Z. Which of these homomorphisms are isomorphisms?
7. Let G be a group and let a ∈ G. Define a map φ : G → G by φ(g) = a−1ga. Is φ a homomorphism?Is φ an isomorphism?
8. Write DR−1 and D−1R in disjoint cycle notation. Can you use these to find a move which changes theorientations of two corner cubies without affecting any other corner cubies?
9. So far, we have only used twists of the 6 faces D, U, L, R, F, and B. Let MR be a clockwise twist(looking at the right face) of the face between the left and right faces. Write MR, MRU, MRU−1, MRU2,and MR−1U2 in disjoint cycle notation. Use these to . . .
(a) . . . find a move in G which cycles 3 edge cubies without affecting any other cubies.(b) . . . find a move in G which changes the orientations of 2 edge cubies without affecting any other
cubies.
Note: Remember that we defined G to be the moves composed of sequences of D,U, L,R,F,B; thatis, G = 〈D,U, L,R,F,B〉. Therefore, MR is not an element of G, so you will have to rewrite your movesto make them elements of G.
10. Is it possible for the Rubik’s cube to be in a configuration where exactly two cubies are in the wrongpositions?
29
9. The Alternating Group
In the previous section, we defined what it meant for an element of Sn to be even or odd. Recall that σ ∈ Snis defined to be even if it can be written as a product of an even number of 2-cycles, and it is defined to beodd if it can be written as a product of an odd number of 2-cycles.
Example 9.1. (1 2)(1 3) is even since it is a product of two 2-cycles. (1 2) is odd since it is a product ofone 2-cycle. ❖
We then proved that an element of Sn is either even or odd, but not both. The tool we used for this wasthe sign homomorphism. Remember that this was a homomorphism ε : Sn → {±1} such that ε(σ) = −1 forany 2-cycle σ. Since the 2-cycles generate Sn, this property characterizes the homomorphism. In fact, theway ε was actually defined is no longer important!
Since ε is a homomorphism, ε(σ) = −1 if σ is odd, and ε(σ) = 1 if σ is even.
Example 9.2. ε((1 2)(1 3)) = 1 and ε((1 2)) = −1. ❖
Example 9.3. ε(1 6 3 4 2) = 1 because (1 6 3 4 2) = (1 6)(1 3)(1 4)(1 2) is even. ❖
Example 9.4. If σ is a k-cycle, then ε(σ) = (−1)k−1. After all, if σ is a k-cycle, then we can writeσ = (a1 a2 . . . ak) = (a1 a2)(a1 a3) · · · (a1 ak). ❖
The product of an even permutation and an odd permutation is odd. The product of two even permutationsor two odd permutations is even. The inverse of an even permutation is even, and the inverse of an oddpermutation is odd. Therefore, we can define a subgroup of Sn consisting of all the even permutations. Thisgroup is called the alternating group and is denoted An.
Example 9.5. If M ∈ G is a face twist (one of D,U, L,R,F,B), then φcube(M) is a product of two 4-cycles.A 4-cycle is odd, so a product of two 4-cycles is even. Therefore, φcube(M) is even. Since the face twistsgenerate all of G, this means that φcube(M) is even for all M ∈ G. That is, φcube(M) ∈ A20 for all M ∈ G.Another way of writing this is to say that im φcube(M) ∈ A20.
Now, φcube(M) = φcorner(M)φedge(M), so either φcorner(M) and φedge(M) are both even, or they are bothodd. That is, φcorner(M) and φedge(M) have the same sign.
Suppose your cube is in the start configuration and you do the move M to it. Then, it ends up in aconfiugration (σ, τ, x, y) where σ = φcorner(M) and τ = φedge(M). Therefore, we have proved that, if(σ, τ, x, y) is a valid configuration, then σ and τ have the same sign. ❖
Since An consists of all of the even elements of Sn, An can also be described as {σ ∈ Sn : ε(σ) = 1}. Thisdefinition can be generalized to any homomorphism.
Definition 9.6. The kernel of a homomorphism φ : G→ H is defined to be {g ∈ G : φ(g) = 1H}, and it isdenoted by kerφ. That is, kerφ is the preimage of 1H in G.
Example 9.7. The kernel of the homomorphism φcube : G→ S20 consists of all moves of the Rubik’s cubewhich do not change the positions of any of the cubies. That is, kerφcube consists of all moves which onlyaffect the orientations, not the positions, of cubies. As you can imagine, this is a useful set to understand:if you have put all the cubies in the right positions, you want to find moves that only affect the orientationsof the cubies. ❖
Theorem 9.8. If G and H are groups and φ : G→ H is a homomorphism, then kerφ is a subgroup of G.
Proof. By Lemma 3.4, it suffices to show that, if g, h ∈ kerφ, then gh−1 ∈ kerφ. So, let g, h ∈ kerφ. Then,
φ(gh−1) = φ(g)φ(h−1) since φ is a homomorphism= φ(g)φ(h)−1 by [PS 7, #5b]
30
= 1H1−1H since g, h ∈ kerφ
= 1H
Therefore, gh−1 ∈ kerφ.
Example 9.9. The alternating group An is the kernel of ε : Sn → {±1}. ❖
Exercises
1. Let σ ∈ S10 be defined by σ = (1 3)(2 4 6 9)(1 4 9). What is ε(σ)? Is σ even or odd?
2. If σ ∈ Sn, prove that σ2 ∈ An.
3. If σ ∈ Sn has cycle type n1, . . . , nr, what is ε(σ)?
4. Prove that An is generated by the set of 3-cycles in Sn.
5. Prove that Sn is generated by (1 2), (2 3), (3 4), . . . , (n− 1 n).
6. Find a move M ∈ G which changes the orientations of the cubies dfr and ulb without affecting anyother corner cubies. What is φcorner(M)? What is a strategy for fixing the orientations of all cornercubies?
31
10. Group Actions
If the Rubik’s cube is some configuration C = (σ, τ, x, y), then doing a move M ∈ G puts the Rubik’s cubein some new configuration. Let’s write this new configuration as C ·M .
Suppose the Rubik’s cube starts in the configuration C. If we do the move M1, the configuration of the cubebecomes C ·M1. If we then do another move M2, the configuration becomes (C ·M1) ·M2. On the other hand,what we have really done is started with the configuration C and applied the move M1M2, so another wayto write the new configuration is C · (M1M2). That is, we have just shown that (C ·M1) ·M2 = C · (M1M2)for all configurations C and all moves M1,M2 ∈ G.
If we do the empty move (the identity element e of G), then the configuration does not change at all, soC · e = C.
This is an example of a mathematical object called a “group action.” Elements of a group (here, the elementsare moves of the Rubik’s cube) affect elements of some set (the set of configurations of the Rubik’s cube).We have actually used group actions already; for instance, to understand Sn, we studied how elements of Snaffected the integers 1, . . . , n.
To give a formal definition, we first need some notation. If S1 and S2 are two sets, then S1 × S2 is the setof ordered pairs (s1, s2) with s1 ∈ S1 and s2 ∈ S2.
Definition 10.1. A (right) group action of a group (G, ∗) on a (non-empty) set A is a map A × G → A(that is, given a ∈ A and g ∈ G, we can produce another element of A, which we write a · g) satisfying thefollowing two properties:
1. (a · g1) · g2 = a · (g1 ∗ g2) for all g1, g2 ∈ G and a ∈ A.2. a · e = a for a ∈ A (here, e is the identity element of G).
This is a right action rather than a left action because we put the elements of the group on the right.
In the first condition, a · g1 ∈ A, so (a · g1) · g2 makes sense. On the other hand, g1g2 ∈ G, so a · (g1g2) alsomakes sense.
When we have a group action of G on a set A, we just say “G acts on A.”
Example 10.2. The group G acts on the set of configurations (σ, τ, x, y) of the Rubik’s cube (we allowboth valid and invalid configurations). ❖
Example 10.3. Sn acts on the set {1, . . . , n}. The group action is defined as follows: given i ∈ {1, . . . , n}and σ ∈ Sn, let i · σ = σ(i). To check that this really is a group action, observe that i · (στ) = (στ)(i) =τ(σ(i)) = τ(i · σ) = (i · σ) · τ and i · 1 = 1(i) = i. ❖
Example 10.4. Sn acts on the set of polynomials in the variables x1, . . . , xn; in fact, we used this actionto prove the existence of the sign homomorphism. Namely, if p(x1, . . . , xn) was a polynomial, we defineda new polynomial pσ by pσ(x1, . . . , xn) = p(xσ(1), . . . , xσ(n)). This was again a polynomial in the variablesx1, . . . , xn. We proved that (pσ)τ = pστ , and it is clear that p1 = p. Thus, if we define p · σ = pσ, we have agroup action. ❖
Example 10.5. The group (Z,+) acts on the set R by a · g = g + a for g ∈ Z and a ∈ R. After all,
(a · g1) · g2 = (a · g1) + g2
= (a+ g1) + g2
= a+ (g1 + g2)= a · (g1 + g2)
for all g1, g2 ∈ Z and a ∈ R. Moreover, a · 0 = 0 + a = 0 for all a ∈ R. ❖
Example 10.6. Often, we are interested in the case when the set A is the group itself. In this case, we say
32
that the group acts on itself. For instance, we can define a group action as follows: for g ∈ G and a ∈ G,define a · g = ag, the normal group multiplication of a and g (check that this defines a group action). Wecall this the action of G on itself by right multiplication. ❖
Definition 10.7. If G acts on a set A, then the orbit of a ∈ A (under this action) is the set {a · g : g ∈ G}.
Example 10.8. G acts on the set of configurations of the Rubik’s cube. The orbit of the start configurationunder this action is exactly the set of valid configurations of the Rubik’s cube. ❖
Example 10.9. In Example 10.5, we said that (Z,+) acts on the set R by a · g = g + 1 for all g ∈ Z anda ∈ R. Thus, the orbit of a is the set {a + g : g ∈ Z}, or the set {. . . , a − 2, a − 1, a, a + 1, a + 2, . . .}. Inparticular, a, a+ 1, a− 1, . . . all have the same orbit. There is a distinct orbit for each a ∈ [0, 1). Therefore,we can think of the set of orbits as the interval [0, 1). However, since the orbit of 0 is the same as the orbitof 1, we could also think of the set of orbits as [0, 1] with 0 and 1 viewed as the same point. One way tovisualize this is to imagine bending the interval [0, 1] around so that 0 and 1 join — this forms a circle! Thus,it is natural to think of the set of orbits of this action as forming a circle. ❖
Definition 10.10. If a group action has only one orbit, we say that the action is transitive (or that thegroup acts transitively).
Example 10.11. G acts on the set of ordered pairs (C1, C2) of different unoriented corner cubies. After all,if C1 and C2 are two different unoriented corner cubies, applying a move M ∈ G sends these corner cubiesto two different corner cubicles C ′1 and C ′2. Then, we can define the group action by (C1, C2) ·M = (C ′1, C
′2).
(Check that this is a group action.) By [PS 3, #5], this action is transitive.
In the same way, G acts on the set of ordered triples (C1, C2, C3) of different unoriented edge cubies. ❖
We often want to prove something about all elements of an orbit (for example, we might want to prove astatement about all valid configurations of the Rubik’s cube). The following lemma can be useful in thesesituations.
Lemma 10.12. Suppose a finite group G acts on a set A, and let S be a set of generators of G. Let P bea property such that the following is true:
Whenever a ∈ A satisfies P and s ∈ S, a · s also satisfies P .
Then, if a0 ∈ A satisfies P , every element in the orbit of a0 also satisfies P .
Proof. Let’s define a new property Q as follows: say g ∈ G satisfies property Q when the following is true:
Whenever a ∈ A satisfies P , a · g also satisfies P .
It suffices to show that every g ∈ G satisfies property Q. After all, that would mean that, if a0 ∈ A satisfiesP , then a0 · g satisfies P for all g ∈ G, which is exactly what we want to show.
By hypothesis, every element of S satisfies property Q. By Proposition 4.9, all we need to show is that, ifg, h ∈ G both satisfy property Q, then gh satisfies property Q. So, suppose g, h ∈ G both satisfy propertyQ. To show that gh also satisfies property Q, we want to show that, if a ∈ A satisfies P , then a · gh alsosatisfies P .
Suppose a ∈ A satisfies P . Since g satisfies property Q, a · g satisfies property P . Since h satisfies propertyQ, (a · g) · h satisfies property P . However, by the definition of a group action, (a · g) · h = a · gh. So, wehave proved that, if a ∈ A satisfies P , then a ·gh satisfies P . That means that gh satisfies property Q, whichfinishes our proof.
In the case of the Rubik’s cube, we will often try to apply the above lemma to the action of the group G onthe set A of configurations. In particular, if we let S = {D,U, L,R,F,B} and a0 be the start configuration,
33
then we can use the lemma to prove things about all valid configurations of the Rubik’s cube.
Exercises
1. Prove that the group (nZ,+) acts on Z by a · g = a+ g for all g ∈ nZ and a ∈ A. What are the orbitsof this action? How many different orbits are there? Does the set of orbits remind you of anything innumber theory?
2. Let G be a group. Prove that G acts on G by a · g = g−1ag for all g ∈ G and a ∈ G. We say that “Gacts on itself by conjugation.”
3. By [PS 10, #2], Sn acts on Sn by τ ·σ = σ−1τσ for all τ ∈ Sn and σ ∈ Sn. What are the orbits of thisaction? (You might want to first try to write out the orbits explicitly for n = 3.)
4. Let R2 be the usual xy-plane, which consists of ordered pairs (x, y) where x, y ∈ R. Prove that thegroup (R,+) acts on R2 by (x, y) · r = (x + r, y) for (x, y) ∈ R2 and r ∈ R. If (x, y) ∈ R2, find theorbit of (x, y). Can you describe the set of orbits geometrically?
5. Let Z2 be the set of ordered pairs (z1, z2) where z1, z2 ∈ Z. We add two ordered pairs as follows:(z1, z2) + (z3, z4) is defined to be (z1 + z3, z2 + z4). Prove that (Z2,+) is a group, and prove that thisgroup acts on R2 by (x, y) · (z1, z2) = (x + z1, y + z2) for all (x, y) ∈ R2 and (z1, z2) ∈ Z2. Can youdescribe the set of orbits geometrically?
6. Let A be the set of ordered triples (C1, C2, C3) where C1, C2, and C3 are different unoriented edgecubies. In class, we explained how G acts on A. What does it mean (in terms of the Rubik’s cube) tosay that this action has only one orbit? Convince yourself that the action really has just one orbit.
7. If C1, C2, and C3 are any three different unoriented edge cubies, prove that there is a move M ∈ Gsuch that M does not affect any corner cubies and φedge(M) = (C1 C2 C3).
8. Suppose your Rubik’s cube is in a valid configuration (e, τ, x, y) (that is, all of the corner cubies are inthe right positions). Prove that τ is even and that there is a move M ∈ G which puts all of the edgecubies in the right positions (without affecting the corner cubies).
34
11. Valid Configurations of the Rubik’s Cube
Now, we will put everything we have learned together to give a characterization of the valid configurationsof the Rubik’s cube.
Theorem 11.1. A configuration (σ, τ, x, y) is valid iff sgnσ = sgn τ ,∑xi ≡ 0 (mod 3), and
∑yi ≡
0 (mod 2).
The rest of this section will be devoted to proving this theorem. First, we will show that, if (σ, τ, x, y) isvalid, then sgnσ = sgn τ ,
∑xi ≡ 0 (mod 3), and
∑yi ≡ 0 (mod 2). In the process, we will prove some
slightly more general facts that will be useful for proving the converse.
Recall that G acts on the set of configurations of the Rubik’s cube. The valid configurations form a singleorbit of this action. So, it makes sense that statements we make about valid configurations can be generalizedto other orbits.
Lemma 11.2. If (σ, τ, x, y) and (σ′, τ ′, x′, y′) are in the same orbit, then (sgnσ)(sgn τ) = (sgnσ′)(sgn τ ′).
Proof. By Lemma 10.12, it suffices to show that, if (σ′, τ ′, x′, y′) = (σ, τ, x, y) ·M where M is one of the 6basic moves, then (sgnσ)(sgn τ) = (sgnσ′)(sgn τ ′). By [PS 8, #3], σ′ = σφcorner(M) and τ ′ = τφedge(M).Therefore, (sgnσ′)(sgn τ ′) = (sgnσ)(sgnφcorner(M))(sgn τ)(sgnφedge(M)). If M is one of the 6 basic moves,then φcorner(M) and φedge(M) are both 4-cycles, so they both have sign −1. Thus, (sgnσ′)(sgn τ ′) =(sgnσ)(sgn τ).
Corollary 11.3. If (σ, τ, x, y) is a valid configuration, then sgnσ = sgn τ .
Proof. This is a direct consequence of Lemma 11.2 since any valid configuration is in the orbit of the startconfiguration (1, 1, 0, 0).
Lemma 11.4. If (σ′, τ ′, x′, y′) is in the same orbit as (σ, τ, x, y), then∑x′i ≡
∑xi (mod 3) and
∑y′i ≡∑
yi (mod 2).
Proof. In light of Proposition 10.12, it suffices to show that, if (σ′, τ ′, x′, y′) = (σ, τ, x, y) ·M where M is oneof the 6 basic moves, then
∑x′i ≡
∑xi (mod 3) and
∑y′i ≡
∑yi (mod 2). You should have done this in
[PS 6, #1]. Here is a table showing what x′ and y′ are if (σ′, τ ′, x′, y′) = (σ, τ, x, y) ·M and M is one of the6 basic moves. In each case, it is easy to check that
∑x′i ≡
∑xi (mod 3) and
∑y′i ≡
∑yi (mod 2).
M x′ and y′
D (x1, x2, x3, x4, x8, x5, x6, x7)(y1, y2, y3, y4, y5, y6, y7, y8, y10, y11, y12, y9)
U (x2, x3, x4, x1, x5, x6, x7, x8)(y4, y1, y2, y3, y5, y6, y7, y8, y9, y10, y11, y12)
R (x1, x7 + 1, x2 + 2, x4, x5, x6, x8 + 2, x3 + 1)(y1, y7, y3, y4, y5, y2, y10, y8, y9, y6, y11, y12)
L (x4 + 2, x2, x3, x5 + 1, x6 + 2, x1 + 1, x7, x8)(y1, y2, y3, y5, y12, y6, y7, y4, y9, y10, y11, y8)
F (x6 + 1, x1 + 2, x3, x4, x5, x7 + 2, x2 + 1, x8)(y1, y2, y8 + 1, y4, y5, y6, y3 + 1, y11 + 1, y9, y10, y7 + 1, y12)
B (x1, x2, x8 + 1, x3 + 2, x4 + 1, x6, x7, x5 + 2)(y6 + 1, y2, y3, y4, y1 + 1, y9 + 1, y7, y8, y5 + 1, y10, y11, y12)
As an example, we’ll see how to find x′ when M is the move R. The cubicles of the right hand face look likethis:
35
u u uf r r r bf r r r bf r r r b
d d d
The cubicles are labeled like this:
2 3
7 8
Therefore, if the Rubik’s cube is in the configuration (σ, τ, x, y), the cubies on the right face are labeled likethis:
x2 x3
x2 + 2 x2 + 1 x3 + 2 x3 + 1
x7 + 1 x7 + 2 x8 + 1 x8 + 2x7 x8
If we rotate this face by 90◦ clockwise, then the cubies look like:
x7 + 1 x2 + 2x7 x7 + 2 x2 + 1 x2
x8 x8 + 1 x3 + 2 x3
x8 + 2 x3 + 1
Thus, x′ = (x1, x7 + 1, x2 + 2, x4, x5, x6, x8 + 2, x3 + 1). So,∑x′i =
∑xi + 6 ≡
∑xi (mod 3).
Corollary 11.5. If (σ, τ, x, y) is a valid configuration, then∑xi ≡ 0 (mod 3) and
∑yi ≡ 0 (mod 2).
Proof. This is a direct consequence of Lemma 11.4 since any valid configuration is in the orbit of the startconfiguration (1, 1, 0, 0).
Thus, we have proved one directon of Theorem 11.1. Now, we will prove the converse. Suppose sgnσ = sgn τ ,∑xi ≡ 0 (mod 3), and
∑yi ≡ 0 (mod 2). We want to show that there is a series of moves which, when applied
to (σ, τ, x, y), gives the start configuration; that is, if the Rubik’s cube is in the configuration (σ, τ, x, y), itcan be solved. The idea of the proof is basically to write down the steps required to solve the Rubik’s cube.Thus, we will prove these four facts:
1. If (σ, τ, x, y) is a configuration such that sgnσ = sgn τ ,∑xi ≡ 0 (mod 3), and
∑yi ≡ 0 (mod 2), then
there is a move M ∈ G such that (σ, τ, x, y) ·M has the form (1, τ ′, x′, y′) with sgn τ ′ = 1,∑x′i ≡ 0
(mod 3), and∑y′i ≡ 0 (mod 2). That is, we can put all the corner cubies in the right positions.
2. If (1, τ, x, y) is a configuration with sgn τ = 1,∑xi ≡ 0 (mod 3), and
∑yi ≡ 0 (mod 2), then there is
a move M ∈ G such that (1, τ, x, y) ·M has the form (1, τ ′, 0, y′) with sgn τ ′ = 1 and∑y′i ≡ 0 (mod
2). That is, we can put all the corner cubies in the right orientations (and positions).
3. If (1, τ, 0, y) is a configuration with sgn τ = 1 and∑yi ≡ 0 (mod 2), then there is a move M ∈ G such
that (1, τ, 0, y) ·M has the form (1, 1, 0, y′) with∑y′i ≡ 0 (mod 2). That is, we can put all the edge
cubies in the right positions (without disturbing the corner cubies).
4. If (1, 1, 0, y) is a configuration with∑yi ≡ 0 (mod 2), then there is a move M ∈ G such that
(1, 1, 0, y) ·M = (1, 1, 0, 0). That is, we can solve the cube!
36
Before proving these, let’s point out a useful fact. Suppose that (σ, τ, x, y) satisfies sgnσ = sgn τ ,∑xi ≡
0 (mod 3), and∑yi ≡ 0 (mod 2). Then, Lemma 11.2 and 11.4 show that, for any (σ′, τ ′, x′, y′) in the same
orbit as (σ, τ, x, y), sgnσ′ = sgn τ ′,∑x′i ≡ 0 (mod 3), and
∑y′i ≡ 0 (mod 2). Thus, for example, in the
first statement above, if we can prove that there is a move M ∈ G such that (σ, τ, x, y) ·M has the form(1, τ ′, x′, y′), it is automatic that sgn τ ′ = 1,
∑x′i ≡ 0 (mod 3), and
∑y′i ≡ 0 (mod 2). Therefore, to finish
the proof of Theorem 11.1, it suffices to prove the following four propositions.
Proposition 11.6. If (σ, τ, x, y) is a configuration such that sgnσ = sgn τ ,∑xi ≡ 0 (mod 3), and
∑yi ≡
0 (mod 2), then the orbit of (σ, τ, x, y) contains some configuration of the form (1, τ ′, x′, y′).
Proposition 11.7. If (1, τ, x, y) is a configuration with sgn τ = 1,∑xi ≡ 0 (mod 3), and
∑yi ≡ 0 (mod 2),
then the orbit of (1, τ, x, y) contains some configuration of the form (1, τ ′, 0, y′).
Proposition 11.8. If (1, τ, 0, y) is a configuration with sgn τ = 1 and∑yi ≡ 0 (mod 2), then the orbit of
(1, τ, 0, y) contains some configuration of the form (1, 1, 0, y).
Proposition 11.9. If (1, 1, 0, y) is a configuration with∑yi ≡ 0 (mod 2), then the orbit of (1, 1, 0, y) con-
tains the start configuration (1, 1, 0, 0).
We will prove these in order. So, we want to first show that we can put all the corner cubies in the rightpositions.
Lemma 11.10. The homomorphism φcorner : G→ S8 is onto.
Proof. By [PS 5, #9], S8 is generated by the set S of 2-cycles in S8. It suffices to show that S ⊂ im φcorner.After all, if S ⊂ im φcorner, then S8 = 〈S〉 ⊂ 〈im φcorner〉 by [PS 4, #2]. By [PS 7, #6], im φcorner is a group,so 〈im φcorner〉 = im φcorner by [PS 4, #1].
So, we want to show that every 2-cycle in S8 is in the image of φcorner. In [PS 6, #2b], you shouldhave found a move which switches just 2 corner cubies and leaves the other corner cubies fixed. Onesuch move is M0 = ([D,R]F)3, which has disjoint cycle decomposition (dbr urb)(dr uf)(br rf)(df lf). Then,φcorner(M0) = (dbr urb). So, we at least know that (dbr urb) lies in the image of φcorner.
Let C1 and C2 be any pair of corner cubies. By [PS 3, #5], there exists a move M ∈ G which sends dbr to C1
and urb to C2. Let σ = φcorner(M). Then, σ(dbr) = C1 and σ(urb) = C2. Since φcorner is a homomorphism,
φcorner(M−1M0M) = φcorner(M)−1φcorner(M0)φcorner(M)= σ−1(dbr urb)σ= (σ(dbr) σ(urb))by [PS 5, #10]= (C1 C2)
Therefore, (C1 C2) ∈ im φcorner, which finishes the proof.
Proof of Proposition 11.6. By Lemma 11.10, there exists a move M ∈ G such that φcorner(M) = σ−1. By[PS 8, #3], (σ, τ, x, y) ·M = (1, τ ′, x′, y′) for some τ ′ ∈ S12, x′ ∈ (Z/3Z)8, and y′ ∈ (Z/2Z)12.
Next, we will prove Proposition 11.7. The basic idea for orienting all of the corner cubies correctly was touse moves which change the orientations of just 2 cubies. First, we must show that such moves exist.
Lemma 11.11. If C1 and C2 are any two corner cubies, there is a move M ∈ G which changes theorientations (but not positions) of C1 and C2 and which does not affect the other corner cubies at all.Moreover, there is such a move M which rotates C1 clockwise and rotates C2 counterclockwise.
Proof. As in the proof of Lemma 11.10, the point is to first find a single move M0 which changes theorientations of 2 cubies and then conjugate M0 to find other moves that change the orientations of 2 cubies.
37
In [PS 6, #3], you should have found such a move; one possibility is M0 = (DR−1)3(D−1R)3, which hasdisjoint cycle decomposition (dfr rdf frd)(drb rbd bdr)(df dr fr ur br db dl). Then, φcorner(M0) = 1 andψcorner(M0) = (dbr rdb brd)(drf rfd fdr). So, if C1 = dbr and C2 = drf, the lemma is true.
Now, we will conjugate this move. By [PS 3, #5], there exists M ∈ G which sends dbr to C1 and drf to C2.Let M ′ = M−1M0M . By applying [PS 5, #10] to find ψcorner(M ′), we see that M ′ changes the orientationsof C1 and C2 and does not affect the other corner cubies. Specifically, M ′ rotates C1 clockwise and rotatesC2 counterclockwise.
Proof of Proposition 11.7. Suppose that the Rubik’s cube is in a configuration where at least two cornercubies C1 and C2 have the wrong orientation. By Lemma 11.11, there is a move which rotates C1 clockwise,rotates C2 counterclockwise, and does not affect the other corner cubies. By applying this move once ortwice, we can ensure that C1 has the correct orientation. Since this move does not affect any corner cubiesbesides C1 and C2, the Rubik’s cube now has one fewer corner cubie with an incorrect orientation. Doingthis repeatedly, we end up with a configuration (1, τ ′, x′, y′) where there is at most one corner cubie withthe incorrect orientation. That is, at least 7 of the x′i are 0. By Lemma 11.4,
∑x′i ≡
∑xi ≡ 0 (mod 3), so
it must be the case that the last x′i is also 0, so the configuration of the Rubik’s cube is (1, τ ′, 0, y′).
Next, we want to prove Proposition 11.9; that is, we want to fix the positions of the edge cubies. Theidea of the proof is very similar to the one we used to prove Proposition 11.7. Recall that, in that case,we first proved that φcorner : G → S8 is onto. In this case, we only want to use moves that don’t affectthe corner cubies, since we have already done a lot of work to get the corner cubies in the right positionsand orientations. Therefore, instead of looking directly at φedge, we will look at the restriction of φedge tokerψcorner (see [PS 8, #5]).
Lemma 11.12. The image of φedge|kerψcorner : kerψcorner → S12 contains A12.
Proof. By [PS 9, #4], A12 is generated by the set of 3-cycles in A12. By the same argument as in the proofof Lemma 11.10, it suffices to show that every 3-cycle is in the image of φedge|kerψcorner . As in the proof ofLemma 11.10, the strategy is to use conjugates of a single move to prove this.
You shoudl have found a move in [PS 8, #9a] that does not affect any corner cubies but cycles 3 edgecubies. One such move is M0 = LR−1U2L−1RB2, which has disjoint cycle decomposition (ub uf db). Then,M0 ∈ kerψcorner, and φedge(M0) = (ub uf db). By [PS 10, #6], if C1, C2, and C3 are any 3 cornercubies, there is a move M of the Rubik’s cube which sends ub to C1, uf to C2, and db to C3. Then,by [PS 5, #10], M ′ = M−1M0M has disjoint cycle decomposition (C1 C2 C3), so M ′ ∈ kerψcorner andφedge(M ′) = (C1 C2 C3). Therefore, (C1 C2 C3) ∈ im φedge|kerψcorner , which completes the proof.
Remark 11.13. In fact, the image of φedge|kerψcorner : kerψcorner → S12 is exactly A12, which you can proveusing Corollary 11.3.
Now, Proposition 11.8 follows directly from Lemma 11.12. (The proof is exactly the same idea as the proofof Proposition 11.6.)
Finally, we must prove Proposition 11.9. This is quite similar to Proposition 11.7; first, we need an analogof Lemma 11.11.
Lemma 11.14. If C1 and C2 are any two edge cubies, there is a move M ∈ G which changes the orientations(but not positions) of C1 and C2 and which does not affect the other cubies at all.
Proof. In [PS 8, #9b], you should have found a move which switches the orientations of 2 edge cubies withoutaffecting any other cubies. One such move is
LR−1FLR−1DLR−1BLR−1ULR−1F−1LR−1D−1LR−1B−1LR−1U−1
(this move is described more easily as (MRU)4(MRU−1)4). Call this move M0; it has disjoint cycle decom-position (fu uf)(bu ub). By [PS 10, #6], G acts transitively on the set of ordered triples (C1, C2, C3) where
38
C1, C2, and C3 are different edge cubies. In particular, if C1 and C2 are any two different edge cubies, thereexists M ∈ G sending uf to C1 and ub to C2. By [PS 5, #10], MM0M
−1 changes the orientations of C1
and C2 and does not affect the other cubies at all.
Now, the argument we used to prove Proposition 11.7 proves Proposition 11.9 as well. This completes theproof of Theorem 11.1.
Remark 11.15. Earlier, we calculated that there were 212388!12! possible configurations of the Rubik’s cube;now, Theorem 11.1 tells us that only 1
12 of those are valid. Of course, this means there are still more than4× 1019 valid configurations, no small number!
39